Quantifying QB Throw Decision-Making in Football

Zachary Pipping, Lou Zhou | Karim Kassam

Motivating Example

Motivating Questions

  • How likely would other QBs make that same decision?
  • Can we quantify how reckless or conservative a QB is?
    • Are QBs being too reckless or too conservative in their throws?
  • Contextualizing touchdown throws / interceptions
    • Would other QBs make the same decision?
  • Look to build a ranking model which determines the most likely receiver at a frame given throw attempt

Data Overview

  • 2025 NFL Big Data Bowl – Weeks 1–9
  • Game and Play Data – Teams, Score, Play Description, Game Context, Play Result, Changes in Win Probability
  • Player Play Data – Statistics for each player for a play
  • Tracking Data - Locations of players and the football at each frame of a play
  • Exclusively looking at throwing plays with an obvious target
    • Removing spikes and throwaways

Current Spacing Tells an Incomplete Story

Speed and Orientation as a Proxy for Future Separation

Methodology

  • Building a ranking algorithm(i.e. XGBoost) to rank the likeliest recipient at a frame - 53.0 \(\pm\) 0.6% accuracy
    • Imputing Features like distance from tracking and event data
    • Previous Work: Deep Learning Approach1, 59.8% accuracy
  • Applying model to quantify recklessness / conservative tendencies of quarterbacks

Feature Set

  • Recipient Features - Distances, Differences in Orientations and Speeds from Top 5 Closest Defenders, Whether the Throw will Result in a First Down, Space Creation1, Receiver Skill
  • Quarterback Features - Distance from Receiver, Movement Vector, Defensive Pressure
  • Game Context - Win Probability, Quarter, Down and Distance, Score Differential

Future Steps

  • Continuing to build feature set to increase model performance
    • Looking to perform comparably with previous deep learning approach
  • Cutting down on Redundant Features and Parameter Tuning
    • Building the simplest model with the strongest predictive power
  • Quantifying Recklessness with Model Outputs

Appendix